UUCF Summer RE 2011
Brain Glitches
Session 1: Confirmation Bias
Bias
What does “bias” mean?
Bias originally referred to a slanted or oblique
line.
Bias is an inclination to present or hold a
perspective at the expense of (possibly equally
valid) alternatives
(easier) An inclination (often against), a
partiality, a prejudice.
Choose 3 adjectives
“George Washington was a ________
president.”
“McDonald’s is a ________ meal.”
“The New England Patriots are a _______
team”.
Cognitive Bias
A cognitive bias is the human tendency to
make systematic errors in certain
circumstances based on cognitive factors
rather than evidence.
(Easier) A bias resulting from the way our
brain works, instead of other factors.
(Easier still) A Brain Glitch
Decision making
What we really ought to do
• Think through all the
evidence
• Use logic
• Ignore irrelevant
information
• Pay attention
• Sort out the meaningful
patterns from the noise
• Change our minds when
there is good reason
What people usually do
• Take mental shortcuts
• Jump to conclusions
• Be confused by irrelevant
information
• Miss important stuff
• Think we see patterns
everywhere
• Stick to our preconceptions,
even when they turn out to
be wrong
Confirmation Bias
The tendency to seek & find
confirmatory evidence for
one’s beliefs, and to ignore
disconfirmatory evidence
OR: “LA LA LA LA I can’t hear you!!!!”
Confirmation Bias
We seek out and pay attention to things that
agree with what we already think.
We tune out and forget those things that
disagree with us.
(Related) We remember the unusual and
forget the routine.
Example
I am the “Rain Goddess of Camping”!
Whenever I go camping, it rains on me. It’s
like the rain loves me and wants to follow me!
It works every time! (Except for that one time
it didn’t, and that one doesn’t count.)
Another Example
My Husband’s brother Pat, when he was a
teenager, said “Whenever I see another driver
do something stupid, it’s always a woman!
And when it’s not, he’s either black or an old
guy!”
Politics
Who’s right and who’s an idiot?
Opinion
Who’s right and who’s an idiot?
Science
Who’s right and who’s an idiot?
Confirmation Bias
“Be careful. People like to be told what they already
know. Remember that. They get uncomfortable when
you tell them new things. New things…well, new things
aren’t what they expect. They like to know that, say, a
dog will bite a man. That is what dogs do. They don’t
want to know that man bites a dog, because the world
is not supposed to happen like that. In short, what
people think they want is news, but what they really
crave is olds…Not news but olds, telling people that
what they think they already know is true.”
Terry Pratchett through the character Lord Vetinari
from his novel, “The Truth: a novel of Discworld
Book Buying Patterns: Amazon.com,
October 2008
The confirmation bias game
http://hosted.xamai.ca/confbias/index.php
The Backfire Effect
The Misconception: When your beliefs are
challenged with facts, you alter your opinions
and incorporate the new information into
your thinking.
The Truth: When your deepest convictions are
challenged by contradictory evidence, your
beliefs get stronger.
Correlation vs.
Causation
In a Gallup poll, surveyors asked, “Do you believe correlation implies
causation?’”
64% of American’s answered “Yes” .
38% replied “No”.
The other 8% were undecided.
Correlation vs Causation
Correlation tells us two variables are related
Types of relationship reflected in correlation
– X causes Y or Y causes X (causal relationship)
– X and Y are caused by a third variable Z (spurious
relationship)
http://www.youtube.com/watch?v=42c7FAnANdk&feature=related
22
Correlation vs. Causation Example
‘‘The correlation between workers’ education
levels and wages is strongly positive”
Does this mean education “causes” higher
wages?
– We don’t know for sure !
Correlation tells us two variables are related
BUT does not tell us why
http://www.youtube.com/watch?v=UNonyq1yhiE
23
Correlation vs. Causation
Possibility 1
– Education improves skills and skilled workers
get better paying jobs
• Education causes wages to 
Possibility 2
– Individuals are born with quality A which is
relevant for success in education and on the job
• Quality (NOT education) causes wages to 
24
Ice-cream sales are strongly
correlated with death from
drowning rates.
Therefore, ice-cream causes
drowning.
• the evidence is the correlation (in yellow):
• "Kids of teen moms are twice as likely not to
graduate than kids whose moms were over
age 22."
• And the conclusion is the text in the top-right:
• "I'm twice as likely not to graduate high
school because you had me as a teen.”
• What potential confounding factors
is the advertisement failing to
consider as alternate causes?
• Ice cream sales and the incidents of polio are
correlated.
 Skirt lengths and stock prices are highly correlated (as
stock prices go up, skirt lengths get shorter).
 The number of cavities in elementary school children
and vocabulary size are strongly correlated
(negatively)
Without proper interpretation,
causation should not be
assumed, or even implied.

Z  X &Y
When can we imply causation?
When controlled experiments are
performed.
Unless data have been gathered by experimental means
and confounding variables have been eliminated,
correlation never implies causation.
Probability
When we don’t understand the probability of
an event happening, we’re more likely to see
it as significant.
Just because a specific event was unlikely does
not mean that it’s significant or part of a
pattern. Unlikely things happen all the time.
Humans trying to emulate random
sequences will almost never place more
than four heads (or tails) in a row.
In a true random generation, the
probability of at least one string of 5 or
more identical outcomes is 0.999 and for a
sequence of 6 it is 0.96!
Miracles = 1 in a million odds
We see & hear things happen about 1/second
30,000 seconds in one 8-hour day
1 million events per month
Most are insignificant
We should expect about 1 miracle to happen,
on average, once a month
Confirmation bias:
we remember the unusual, forget the usual
Death Dreams
5 dreams/day = 1,825 dreams/year
1/10 remembered dreams = 182.5/year
295 million Americans = 53.8 billion remembered
dreams/year
Each of us knows about 150 people fairly well
Network grid of 44.3 billion personal relationships
Annual U.S. death rate = .008 = 2.6 million/year
Inevitable that some of those 53.8 billion
remembered dreams will be about some of these
2.6 million deaths among the 295 million
Americans and their 44.3 billion relationships.
It would be a miracle, in fact, if some death
premonition dreams did not come true
Birthdays
Two people in your classroom have the
same birthday. That seems like an unlikely
event.
What are the odds, really?
The Birthday Probability Game
The odds of
getting 2
people with
the same
birthday is
better than
50% with
only 23
people.
FALSE POSITIVES
One reason we make mistakes in the first place
A Type I error, or a false positive, is
believing a pattern is real when it
is not (finding a nonexistent
pattern).
A Type II error, or a false negative,
is not believing a pattern is real
when it is (not recognizing a real
pattern).
A Type I error: believe that the
rustle in the grass is a dangerous
predator when it is just the wind
(low cost).
A Type II error: believe that the
rustle in the grass is just the wind
when it is a dangerous predator
(high cost).
Rustle in the Grass
Really
There’s no tiger
There’s a tiger!
I think
Help, it’s a tiger!!
Run!!!
OK, I ran away when I
didn’t need to. No big.
Correct answer! Yay, I
didn’t get eaten!
(Type I error)
RAWR!
Ho, Hum, Just the
wind. Don’t panic,
there’s no tiger.
Correct Answer. Yawn.
(Type II error)
The Downside of False Positives
Suppose that about one in a thousand people
has the dreaded lethal disease creeping
uvulitis, and doesn’t even know it. Suppose
we have a test that is 99% accurate in
diagnosing this dread ailment.
What happens if we test a whole bunch of
people for the disease?
Creeping Uvulitis, we test 100,100 people, 100 of them are sick
Really
All Clear
100,000 people
Disease
100 total
Disease
1,000 people get the
bad news that they’re
sick, when they’re not.
(Type I error)
Correct answer! 99
people are treated and
don’t die.
All Clear
Correct Answer. 99,000 One guy is told he’s
people get the good
fine. He dies.
news that they’re fine. (Type II error)
99% accurate
Test shows
The results
100,100 people are tested with a 99%
accurate test.
1,099 get results that say they are sick. 9% of
these people actually are.
Understand probabilities before you jump to
conclusions.